Auxiliary diagnosis of heterogeneous data of Parkinson’s disease based on improved convolution neural network

Abstract

Parkinson’s disease (PD) is a kind of nervous system degenerative disease frequently occurring in the elderly over sixty years old. With the development of imaging technology, medical imaging has played a certain role in the diagnosis of Parkinson’s disease. The aim of the paper is to the diagnosis of Parkinson’s disease through deep learning. This paper selects the T2-MRI(T2-Magnetic Resonance Imaging) image and clinical data to diagnose Parkinson’s disease and integrates the heterogeneous data into the improved convolution neural network. In this paper, convolution neural network is added to the Gabor filter to make the whole convolution neural network have better effect; the activation function is improved and adjusted, which means the traditional sigmoid function and the tanh function are discarded, and the Relu activation function is used to improve the neural network. It is proved by experiments that the heterogeneous data diagnosis of T2-MRI image and clinical data (the accuracy is 77.9%) is better than the simple image data diagnosis (the accuracy is71.2%). For the same data, the improved convolution neural network is superior to the traditional network (the accuracy is 64.5%).